Managers navigate AI task shifts in workforce workflows | Managers at Work

Listen to this article

“Like many organizations, we’re slowly implementing AI tools into our processes, but we have no idea yet how our jobs are going to change as a result of all this. It feels like chaos at times.”

With the rapid arrival of AI, how are jobs changing? How should they change? What should new work processes and new workloads look like? How well will AI improve productivity and efficiency? Many managers and team leaders are wrestling with these questions as they look at workflows, processes, tasks and decisions that get made on a day-to-day basis.

There are so many outstanding questions. In a report released in February this year, researchers at EY described the problem this way: “AI has entered the workforce far faster than the structures, roles and cultures can absorb it,” researchers said.

Citing 2023 research by the World Economic Forum, the EY report noted that some 75 percent of firms plan to adopt AI within five years, “yet fewer than half of them have redesigned workflows or roles around it.”

Discussions about AI to date have been more about the loss and potential loss of jobs than how existing jobs will be changing. That’s what researchers at McKinsey noted in another report released earlier this year. “The dominant narrative around AI still frames the debate in terms of jobs gained versus jobs lost.”

“That framing is too narrow,” researchers said. “What is changing fastest is the content of work – the tasks that people perform and the skills they apply.”

The report, released in January by McKinsey, showed that some 90 percent of companies reported investing in AI but fewer than 40 percent are seeing “meaningful impact” on the bottom line. This gap is occurring because “many organizations are applying AI to individual tasks, rather than redesigning entire processes or workflows,” researchers said.

As McKinsey researchers put it, workflows – “the connected sequences of activities that deliver outcomes” – were built in another era long before AI. “Layering a chatbot or automation tool onto those legacy processes yields incremental gains at best,” they wrote. “The real productivity unlock comes from reimagining workflows so people, agents and robots each do what they do best to get work done.”

Indeed, experts we talked with agreed that for AI to really improve productivity, the focus should be on workflow and tasks. But they took different approaches to how that should be done – how to conduct these analyses.

Ravi Teja Surampudi, senior manager of GTM Digital Customer Experience at Workday, said that in this evaluation process, he’s learned not to focus on job roles or titles, but instead on the flow of the work. “We started by breaking down roles into tasks and identifying where AI can accelerate execution versus where human review and judgment is still necessary,” he says. “I feel that this is really important because if you skip that step, you end up layering AI on top of existing inefficiencies.”

The company has also changed how it sees AI agents. Instead of being tools, the agents have become part of the engineering workforce, he says. “We give the AI agent a set role similar to a software engineer, for instance, providing code generation, test creation, analysis and operational support, with human engineers concentrating on review, integration and judgment. This has significantly changed the way we consider capacity planning and team design.”

Surampudi noted that he is spending a significant amount of time investing in “guardrails.” “We are creating prompt standards, output validation and escalation paths to make AI usable at scale, especially in customer-facing or high-risk environments.”

Triparna Chakraborty, Human Resources business partner at Credo in San Jose, Calif., agreed that focusing on tasks, not tools, paves the way for more innovation.

Chakraborty, who specializes in workforce planning for AI-augmented teams, says she works with leaders to break every role down into three buckets: tasks that AI can own completely; tasks where AI assists but a human makes decisions, and tasks that should stay fully human.

Tasks where AI assists but a human decides can include customer service. “We can have an AI agent answering most asked questions, but for anything more complicated, a human’s empathy and judgment is critical.” And tasks that should stay fully human should include lab work as an example.

“That simple exercise changes the entire conversation from ‘replacing people’ to ‘where do we want people spending their brainpower?’ and ‘which skills are most valuable to us?’ versus “what tactical work can we handover to AI?’”

Before redesigning jobs and functions, Chakraborty suggests taking several steps:

The first is to audit for ‘invisible pain work’ – what she calls “shadow” work. “Every team has tasks that don’t show up in a job description but hold everything together, such as context-switching between tools, reformatting data between systems, chasing approvals,” she says. “That’s where AI support could save the most time and where your team feels the relief the fastest.”

The second tip is to do the redesign in a 90-day sprint. Sometimes team leaders try to transform an entire role overnight and that creates chaos and panic, she says. “The trick is to treat it like a project to be managed in sprints and use the same project management principles.”

You could, for example, select one workflow per quarter, pilot the AI-assisted version with two or three team members, gather feedback, then expand it, she says. “You learn things and that feeds back into the process. People need time to build confidence with new ways of working.”

She also advises integrating AI slowly and carefully creating frameworks and guidelines to go with it. “Frameworks around AI output need to be verified and guidelines around how to scale to the larger organization need to be developed. “You are working in a hybrid model where humans and AI work together – this needs to be acknowledged and guardrails around that dynamic need to be set up.”

And finally, Chakraborty suggests updating job descriptions. “If you have added AI into someone’s workflow but their job description still reads the same, you have created an issue around trust. “People start wondering, ‘if my job looks the same on paper, does leadership even understand what I’m doing now?”

So take the time to update the job description, update the performance metrics, and have a real conversation about how the employee’s growth path has changed, she says. “At the end of the day AI is a tool. So my recommendation is always to treat job redesign as a people conversation first and a technology conversation second.”

As you examine roles and responsibilities, you need to map out decision-making, says Ryoji Morii, founder of Insynergy in Japan, an independent consulting practice focused on decision-making in organizational transformation.

He suggests creating a “Decision Level Map” that will map out who will be responsible for  decisions that will be made before and after the changes resulting from implementing AI. You need to decide who will be accountable for the decisions that will be made, he says, and teams need clear visibility on decision points.

“They need to know when they are submitting input to an automated system and when they need a human to act,” he says.

Managers at Work is a monthly column exploring the issues and challenges facing managers. Contact Kathleen Driscoll with questions or comments by email at [email protected]